Identification of residues Asn89, Ile90, and Val107 of the factor IXa second epidermal growth factor domain that are essential for the assembly of the factor X-activating complex on activated platelets.

Activated platelets promote intrinsic factor X-activating complex assembly by presenting high affinity, saturable binding sites for factor IXa mediated by two disulfide-constrained loop structures (loop 1, Cys88-Cys99; loop 2, Cys95-Cys109) within the second epidermal growth factor (EGF2) domain. To identify amino acids essential for factor X activation complex assembly, recombinant factor IXa point mutants in loop 1 (N89A, I90A, K91A, and R94A) and loop 2 (D104A, N105A, and V107A) were prepared. All seven mutants were similar to the native factor IXa by SDS-PAGE, active site titration, and content of gamma-carboxyglutamic acid residues. Kinetic constants obtained by either titrating factor X or factor VIIIa on SFLLRN-activated platelets or phospholipid vesicles revealed near normal values of Km(app) and Kd(app)FVIIIa for all mutants, indicating normal substrate and cofactor binding. In a factor Xa generation assay in the presence of activated platelets and cofactor factor VIIIa, compared with native factor IXa (Kd(app)FIXa approximately 1.1 nm, Vmax approximately 12 nm min(-1)), N89A displayed an increase of approximately 20-fold in Kd(app)FIXa and a decrease of approximately 20-fold in Vmax; I90A had an increase of approximately 5-fold in Kd(app)FIXa and approximately 10-fold decrease in Vmax; and V107A had an increase of approximately 3-fold in Kd(app)FIXa and approximately 4-fold decrease in Vmax. We conclude that residues Asn89, Ile90, and Val107 within loops 1 and 2 (Cys88-Cys109) of the EGF2 domain of factor IXa are essential for normal interactions with the platelet surface and for the assembly of the factor X-activating complex on activated platelets.

activated by either FXIa or tissue factor/FVIIa (TF/FVIIa) to form FIXa, a heterodimer consisting of a catalytic-domaincontaining heavy chain (ϳ28 kDa) and light chain (ϳ18 kDa). FIXa exhibits negligible activity toward its physiological substrate FX unless it is in complex with the surface (activated platelets or negatively charged phospholipids) and a non-enzymatic cofactor FVIIIa via a mechanism that remains to be fully understood. The FIXa heavy chain contains the serine protease active site that catalyzes the activation of the macromolecular substrate FX in a cofactor-dependent manner (1). The light chain of FIXa, which consists of a ␥-carboxylated glutamic acid (Gla)-containing module and two modules that are highly homologous to epidermal growth factor (EGF), also participates in the assembly of the FX-activating complex by interacting with both the surface and the cofactor.
Previous studies from our laboratory have demonstrated that the zymogen, FIX, binds to a discrete number (n ϳ 250 sites per platelet) of receptors on the surface of activated platelets (K d ϳ 2.5 nM) that can also be occupied by the enzyme, FIXa, and are mediated by residues Gly 4 -Gln 11 within the Gla domain (2)(3)(4)(5)(6). In addition, FIXa, but not FIX, can bind to a site (n ϳ 250 sites per platelet, K d ϳ 0.5 nM) on activated platelets, mediated by residues 88 -109 (disulfide constrained loops 1 and 2) but not by residues 110 -124 (loop 3) within the EGF2 domain (2)(3)(4)(5)(6). Occupancy of these binding sites is closely correlated with optimal rate enhancements of FX activation (Ͼ2 ϫ 10 8 -fold) in the presence of FVIIIa, emphasizing the physiological significance of platelet-receptor-mediated coagulation complex assembly (2,6,7). Recently, however, alanine-scanning mutagenesis studies of residues within the EGF2 domain of FIX have identified residues Asn 89 -Gly 93 that were implicated as critical for binding of FVIIIa (8). Because it is highly unlikely that the same subdomain of the FIXa EGF2 domain could be important for interactions with both platelet receptors and with FVIIIa, we have conducted the present studies, which are designed to identify specific amino acids essential either for binding to activated platelet receptors or to the cofactor, FVIIIa. "Candidate" residues, most likely to be involved in mediating FX activation on the surface of activated platelets, were selected for alanine-scanning mutagenesis on the basis of: 1) surface exposure of amino acid side chains by x-ray crystallography; 2) conservation among species; and 3) dissimilarity with those in a homologous protein FVII, which does not participate in the assembly of the intrinsic FX-activating complex and does not bind to activated platelets with high affinity (9). Based on these criteria, we selected and prepared recombinant FIXa point mutants in loop 1 (N89A, I90A, K91A, and R94A) and loop 2 (D104A, N105A, and V107A) of the EGF2 domain to study the contributions of these residues to FX activation complex assembly on activated platelets (Fig. 1). The present results support the conclusion that residues Asn 89 , Ile 90 , and Val 107 within loops 1 and 2 (Cys 88 -Cys 109 ) of the EGF2 domain of FIXa are essential for normal interactions with the platelet surface and for the assembly of the FX-activating complex on activated platelets.
HEPES, Tris, fatty acid-free bovine serum albumin (BSA), heparin from porcine intestinal mucosa, benzamidine, and other reagents were purchased from Sigma. Human FIX, human FX, and human antithrombin III (ATIII) were purchased from Enzyme Research Laboratories (South Bend, IN). The FX preparation, obtained as a lyophilized powder, was dissolved in sterile water and dialyzed against HEPES-Tyrodes buffer (HT) before being used in experiments. Human FXIa was purchased from Hematologic Technologies (Essex Junction, VT). High purity recombinant human FVIII was obtained as a generous gift from Baxter Healthcare Corp. (Duarte, CA). Thrombin was purchased from Sigma, the chromogenic substrate S-2765 (N-␣-benzyloxycarbonyl-Darginylglycyl-L-arginine-para-nitroanalidedihydrochloride) was from Dia Pharma Group (Stockholm, Sweden), and bovine brain PS and L-␣-dioleoyl-PC were from Avanti Polar Lipids (Birmingham, AL). The thrombin receptor hexapeptide SFLLRN-amide was synthesized using ((9-fluorenyl)methoxy)-carbonyl (FMOC) chemistry on an Applied Biosystems 430A synthesizer, and by reverse phase HPLC was purified to Ͼ99.9% homogeneity.
In Vitro Mutagenesis and Mutant FIX Expression-Alanine point mutations were generated by using the PCR-based method, FIX mutant proteins were expressed in HEK293 cells, and the highest expressing clones were identified as previously described (8). All wild type and mutant proteins were expressed at similar levels.
Purification of FIX Proteins-Confluent cells were washed with phosphate-buffered saline twice and then switched to serum-free medium supplemented with vitamin K for 24 h. Fresh serum-free medium was added, and conditioned medium was collected every 72 h. The conditioned medium was supplemented with 5 mM benzamidine and 5 mM EDTA and filtered through a cellulose acetate filter (0.45-m pore size) to remove cell debris. The medium was stored at 4°C until the day of purification. Purification of recombinant FIX using ion exchange chromatography on Q-Sepharose was as described (10). For 500 ml of conditioned serum-free medium, a 1-ml slurry of Q-Sepharose beads was used. The Q-Sepharose beads were first thoroughly equilibrated in the equilibration buffer (TBS plus 5 mM benzamidine plus 5 mM EDTA), and then packed in the chromatography column. The conditioned serum-free medium was then chromatographed. The column was washed with 30 column volumes (ϳ60 ml) of equilibration buffer and then further washed with 20 column volumes (ϳ40 ml) of TBS supplemented with 2 mM benzamidine to remove the EDTA. Recombinant FIX proteins were then eluted with TBS 2 mM benzamidine 5 mM CaCl 2 . Fractions were concentrated to ϳ3 ml using Centricon Plus-20 (10,000 molecular weight cut-off, Millipore, Bedford, MA) and dialyzed in TBS buffer.
Protein Concentrations-The concentration of FIX proteins were initially determined using the BCA assay (11) and then were corrected utilizing the results of active site titration as described below.
Activation of FIX Proteins by FXIa-FIX proteins were activated to their active enzymatic forms as follows: FIX proteins (1 M) were diluted in HT supplemented with 5 mM CaCl 2 . FXIa (5 nM) was added at a 1/200 molar ratio, and the reactions were incubated at 37°C for 90 min. Zymogen activation rates of wild type and mutant proteins were similar. Complete activation was confirmed by SDS-PAGE/silver staining and by active site titration with ATIII.
Active Site Titration of FIXa Proteins-To assay the active-site concentration of mutant and wild type proteins, 10 l of FIXa protein (100 nM) was incubated with 10 l of ATIII dilutions (0 -100 nM) in HT (15 mM HEPES,126 mM NaCl, 2.7 mM KCl,1 mM MgCl 2 , 375 M NaH 2 PO 4 , and 5.6 mM glucose, pH 7.2) supplemented with BSA (1 mg/ml), heparin (20 g/ml), and CaCl 2 (5 mM) at 37°C for 15 min, and the reaction was diluted by adding HT buffer to 200 l. Residual FIXa activity was examined by assaying FXa generation activity in the presence of FVIIIa (5 units/ml), FX (400 nM), and L-␣-dioleoyl-PC-PS vesicles (20 M). The FXa generation reaction was allowed to proceed at 37°C for 2 min and was stopped by addition of 50 l of stopping buffer (50 mM HEPES (pH 8.1), 175 mM NaCl, and 20 mM EDTA). The amount of FXa generated was determined by using its chromogenic substrate S2765 as described before (3).
Determination of K mapp and V max in the Presence of Saturating FX-An FX activation vessel, containing 1 nM FIXa, 5 units/ml FVIIIa, and SFLLRN-activated platelets (5 ϫ 10 7 platelets/ml) or extruded PC:PS (mol:mol ϭ 3:1, total concentration ϭ 2 M) vesicles in HT buffer supplemented with BSA (2 mg/ml) and 5 mM CaCl 2 , was used. The reaction was initiated by the addition of FX to the indicated concentration and allowed to proceed for 2 min at 37°C. Then 10 mM EDTA was added to stop the reaction, and the FXa generation was measured as indicated above. 8 )-FVIIIa at indicated concentrations was titrated into the FX activation vessel containing 0.5 nM FIXa and SFLLRN-activated platelets (5 ϫ 10 7 platelets/ml) or extruded PC:PS (mol:mol ϭ 3:1, total concentration ϭ 500 nM) vesicles in HT buffer supplemented with BSA (2 mg/ml) and 5 mM CaCl 2 . The reaction was initiated by the addition of 250 nM FX and allowed to proceed for 2 min at 37°C. Then 10 mM EDTA was added to stop the reaction. The amount of FXa generated was determined as above. V max 8 was defined as the maximum velocity of FXa generation at a saturating concentration of FVIIIa under the experimental conditions described. K d(app)FVIIIa was defined as the concentration of FVIIIa required to achieve half-maximal rates of FX activation. (V max   9 )-FIXa at indicated concentrations was titrated into the FX activation vessel containing 5 units/ml FVIIIa and SFLLRN-activated platelets (5 ϫ 10 7 platelets/ml) in HT buffer supplemented with BSA (2 mg/ml) and 5 mM CaCl 2 . The reaction was initiated by the addition of 250 nM FX and allowed to proceed for 2 min at 37°C. Then 10 mM EDTA was added to stop the reaction. The amount of FXa generated was determined as above. V max 9 was defined as the maximum velocity of FXa generation at a saturating concentration of FIXa under the experimental conditions described. K d(app)FIXa was equivalent to the concentration of FIXa required to achieve half-maximal rates of FXa generation under the experimental conditions described.

Determination of K dappFIXa and FX Activation Velocity in the Presence of Saturating FIXa
Calculation of k cat -Turnover number (k cat ) is defined as moles of FXa generated per mole of surface-bound FIXa per second. Moles of FXa generated was calculated from the V max (nanomolar FXa/min), and moles of surface-bound FIXa was calculated from the determined K d values for the given set of reaction conditions and the input FIXa protein concentration. FIXa-bound concentration (nanomolar) are calculated from Equation 1, where B max (nM) for native FIXa protein was determined from the platelet concentration and the previously determined stoichiometry (600 sites/platelet) for FIXaN (37). B max (nanomolar) for FIXa alanine mutants were derived from kinetic FIXa binding experiments (see Table IV and Fig. 4) by multiplying the B max (nanomolar) value for FIXaN by the -fold differences in V max 9 . Data Analysis-FXa generation rates from all reactions described here were fitted to a hyperbolic curve using a non-linear least squares fit as previously standardized (3).
were derived using KaleidaGraph Software as previously described. Statistical analysis was carried out using analysis of variance followed by pair-wise comparisons with the Bonferroni adjustment procedure for multiple comparison maintaining an experiment-wise Type 1 error level of 0.05 (13) as previously described (3).

RESULTS
Characterization of Proteins-FIX EGF2 alanine-scanning "candidate" mutants expressed in HEK293 cells and secreted in the serum-free medium were purified using Q-Sepharose-dependent pseudoaffinity chromatography as described under "Materials and Methods." Co-migration of each of the recombinant FIX proteins with normal plasma-derived FIX (FIX NP ) by SDS-PAGE (data not shown) suggested normal translation and post-translational modification. Additionally, each of the recombinant FIX proteins was found to have the expected number of ␥-carboxyglutamate (Gla) residues (10.9 -13.2 mol of Gla/mol of protein) (Table I).
Earlier time-course solution-phase FIX activation results have shown that FXIa activates all the seven FIX alanine mutants at normal rates compared with native FIX (8), and each of the FIX proteins was fully activated by FXIa in solution. Gelcode Blue-stained gels of SDS-PAGE displayed complete disappearance of the zymogen band (ϳ70 kDa) and appearance of heavy (ϳ28 kDa) and light chains (ϳ18 kDa) of FIXa for all the recombinant FIXa proteins (data not shown). Active site titration with ATIII also showed complete activation of all FIX proteins. The FIXa concentrations of wild type and mutant proteins used in active site titration were ϳ100 nM, determined from A 280 and BCA assays. The active site concentrations exhibited for the recombinant FIXa proteins were 84 -120 nM (Table I).
F-X Titrations-In preliminary screening studies, all the seven "candidate" alanine mutants displayed decreased FXa generation activities on phospholipids. In the present studies, FX activation activities of these FIXa mutants were thoroughly examined on both SFLLRN-activated platelets and PC:PS (3:1) vesicles, respectively (Fig. 2). In these assays, kinetic parameters of the plasma-derived FIXa (FIXa) were used as positive controls, although recombinant FIXa (rFIXawt) was also included in experiments and showed indistinguishable activities compared with native FIXa. Presented in Table II are the apparent Michaelis-Menten kinetic parameters for the FIXa proteins on both surfaces. Regarding activated platelets on the physiological cell membrane, all seven Ala mutants exhibited reduced V max values in FXa generation when compared with native FIXa (V max ϳ 4.05 nM FXa/min) (Table II and Fig. 2). In particular, rFIXa89A (ϳ1% V max ), rFIXa90A (ϳ3% V max ), and rFIXa107A (ϳ10% V max ) displayed drastically impaired FX activation activities (rescaled in Fig. 2B). However, no significant increase in K m(app) values was observed for most of the chimeras, with the exception of an increase of ϳ3-fold in the case of rFIXa89A. This suggests that the deficiency in FXa generation activity does not result from defective FX association with the enzymatic complex. Substrate titrations on artificial surface, extruded phospholipids (L-␣-dioleoylPC:PS ϭ 3:1) (regression curves not shown), revealed that rFIXa91A, rFIXa94A, rFIXa104A, and rFIXa105A displayed normal V max values compared with native FIXa (V max ϳ 25.1 nM FXa/min) although the K m(app) values were slightly increased (Table II). The three other Ala mutants, rFIXa89A, rFIXa90A, and rFIXa107A displayed significantly decreased V max values, whereas K m(app) values for rFIXa90A and rFIXa107A were not increased. Kinetic values for rFIXa89A on phospholipids were not determined because saturable enzyme kinetic curves were not observed (Table II).
FVIIIa Titrations-FVIIIa was titrated in the presence of either activated platelets or phospholipids to examine FVIIIa interactions with the FIXa chimeras ( Fig. 3 and Table III). For SFLLRN-activated platelets, the maximal velocity of FX activation at saturating FVIIIa concentration was drastically decreased for rFIXa89A (ϳ1% V max ), rFIXa90A (ϳ10% V max ), and rFIXa107A (ϳ20% V max ) compared with native FIXa. However, the effective concentration at half -maximal velocity (K d(app)FVIIIa ) remained the same for all the FIXa proteins in the presence of either activated platelets or phospholipid vesicles. No significant change in K d(app)FVIIIa values was observed for any of the Ala mutants (Table III), indicating normal incorporation of the cofactor into the enzymatic complex on surfaces.
FIXa Titrations-Examination of platelet-mediated FX activation at increasing concentrations of FX or FVIIIa suggested that the deficiency in FX activation complex assembly on platelets observed with the Ala mutant proteins was not due to defective substrate or cofactor incorporation. Enzyme titrations were carried out to assess the platelet binding capacities of the mutants. The values of maximal velocity at saturating concentrations of the enzyme were significantly reduced for rFIXa89A (ϳ4.7% V max ), rFIXa90A (ϳ11.8% V max ), and rFIXa107A (ϳ25% V max ) compared with native FIXa (titration curves rescaled in Fig. 4B), and the binding affinities indicated by K d(app) values were severely impaired for these proteins (Table  IV). Specifically, the K d(app) values for rFIXa89A were ϳ50-fold increased, whereas rIXa90A displayed ϳ12-fold increase and rFIXa107A displayed ϳ8-fold increase in K d(app)FIXa for activated platelets. Results similar to those obtained with activated platelets (Table IV) were observed when the FIXa mutants were titrated on phospholipid vesicles (data not shown). The results indicated that enzyme incorporation was severely defective for the mutants rFIXa89A, rFIXa90A, and rFIXa107A, which have consistently exhibited severely impaired FX-activating velocities in all the experiments carried

EGF2 Domain of FIXa in FX Activation on Platelets
out (Figs. 2-4), and that the surface binding deficit results in the decreased values of V max for FX activation observed in the kinetic assays.
Coagulation Assays-Activated partial thromboplastin time assays for the seven "candidate" alanine mutants were previously reported (8) and are listed in Table IV, which confirms the physiological function of the interrupted surface binding property of the FIX mutants in mediating clotting. The three mutants N89A (0.5%), I90A (7%), and V107A (0.5%) exhibited the lowest clotting activities. DISCUSSION To identify essential residues in the disulfide-constrained loops 1 and 2 of the EGF2 domain (Cys 88 -Cys 109 ) that are important for platelet-mediated FIXa/FVIIIa complex assembly, we utilized the primary amino acid sequence alignment within the disulfide-constrained loops 1 and 2 of the EGF2 domain ( Fig. 1) to identify "candidate" residues that are highly conserved among species and different from those in FVII, because residues in FVII molecules have been shown to be ineffective in mediating platelet surface-mediated FX activation complex assembly (2,6). Recombinant FIXa point mutants in loop 1 (N89A, I90A, K91A, and R94A) and loop 2 (D104A, N105A, and V107A) were prepared to study the contributions of these residues to FX activation complex assembly on activated platelets. All seven mutants were similar to native FIXa by SDS-PAGE (both reduced and non-reduced); active site titration, which confirmed the presence of one intact active site per FIXa mutant molecule; and, content of ␥-carboxyglutamic acid residues. Kinetic constants obtained by titrating FX on SFLLRN-activated platelets in the presence of FVIIIa revealed near normal values of K m(app) for all mutants, indicating normal substrate binding. It should be emphasized that, whereas the observed K m values in substrate (FX) titration experiments are indicative of the affinity of substrate incorporation, K m is not the true substrate dissociation constant (K d or K s ). Rather, the reaction rate contributes to K m (K m ϭ K d ϩ k cat /k on ), which may explain why when the reaction velocity is extremely slow for the FIXa mutants, the K m(app) values were generally decreased (Table II). However, because decreased K m(app) values suggest even tighter substrate association to the enzyme complex, we conclude that the substrate incorporation was not interrupted by any of the mutations with the possible exception of the mutant N89A. Even though the Asn 89 alanine mutant displayed a slightly interrupted substrate incorporation (Table  II), the severely impaired velocity (V max ) in substrate, cofactor, and enzyme titrations (Tables II-IV) for this mutant coincided with an increase of ϳ50-fold in K d(app) for FIXa, and a similar K d(app)FVIIIa value compared with native FIXa, suggesting that enzyme binding to activated platelets or to phospholipids was  2. Determination of K m(app) and V max on SFLLRN-activated platelets. FIXa proteins were diluted to 1 nM in HT buffer containing SFLLRN (5 M)-activated platelets (5 ϫ 10 7 /ml). FVIIIa was added to 5 units/ml, and FX was added to the indicated concentration. After 2 min, the reactions were terminated by addition of EDTA to 10 Table IV), these "corrected" k cat values were all either normal (or slightly increased in the case of the rFIXa89A mutant). This result can be interpreted as demonstrating that the only defect arising from these mutations is a consequence of a platelet (or phospholipid)-binding defect and that the catalytic activity of the mutant enzymes is normal. We conclude that residues Asn 89 , Ile 90 , and Val 107 within loops 1 and 2 (Cys 88 -Cys 109 ) of the EGF2 domain of FIXa are essential for normal interactions with the platelet surface and for the assembly of the FX activating complex on activated platelets. When the locations of these three essential amino acids were displayed using the human FIXa crystal structure (PDB: 1RFN, not shown) they form a planar surface, which the present studies suggest forms a site utilized for binding to activated platelets, consisting of two hydrophobic residues (Ile and Val) and an uncharged polar residue (Asn) that can participate in hydrogen bonding. Extensive studies have been carried out to identify regions in FIXa that are essential in surface (14 -21) or cofactor interaction (1,(22)(23)(24)(25)(26)(27)(28)(29)(30)(31). Regions 301-303 and 333-339 in the FIXa catalytic domain are implicated as FVIIIa-interactive sites, because mutations in these regions in recombinant FIXa chimeric proteins had deleterious effects on FX activation with significantly increased EC 50VIIIa values (26), results supported by direct equilibrium binding studies (32). The non-catalytic light chain (containing the Gla and two EGF domains) of FIXa is also implicated in interaction with FVIIIa (33) from studies using fluorescence anisotropy to assess the effect of FIXa fragments on FVIIIa reconstitution. The Gla domain of FIXa appears to interact directly with FVIIIa in the FX activating complex as suggested by cross-linking experiments, implicating Phe 26 and perhaps Val 46 but not Phe 9 (34). The salt bridge between Glu 78 within the EGF1 domain and Arg 94 within the EGF2 domain is also shown to contribute to stimulation by FVIIIa in FX activation (24) by maintaining FIXa structure, although an FIX antibody directed to the C terminus of EGF1 domain had a marginally inhibitory effect on FX activation (35). The connecting segment between the two EGF domains (Leu 84 -Thr 87 ), but not segment Asn 89 -Lys 91 in the EGF2 domain, appeared to mediate stimulation of FX activation by

EGF2 Domain of FIXa in FX Activation on Platelets
FVIIIa. However, this contribution is considered to be structural rather than direct, because the segment FIXa variants showed normal FVIIIa binding in surface plasmon resonance experiments (36). Kinetic studies using recombinant FIXa chimeras have consistently shown that residues in the EGF2 domain do not mediate FVIIIa association suggested by both normal kinetic stimulation by FVIIIa and normal K d(app)FVIIIa values (2)(3)(4)6). Thus, the contribution of the non-catalytic domain of FIXa to its interaction with FVIIIa appears to be mainly structural, i.e. to promote optimal positioning of the catalytic domain of FIXa to interact with FVIIIa (31). On the other hand, extensive platelet binding combined with FXa generation studies have proven that the light chain of FIXa mediates platelet binding, which leads to FX-activating complex assembly (2, 3, 5-7, 14 -18, 20, 37-41). These studies demonstrate that the platelet-bound (not fluid-phase) FIXa is the physiologically functional enzyme and that surface binding is the driving force for FIXa⅐FVIIIa complex assembly. It is hypothesized that once assembled on the activated platelet or phospholipid membrane, the non-catalytic domain contributes structurally to bring the FIXa catalytic domain in contact with FVIIIa, which alters the local conformation of the FIXa active site (42) and increases its capacity to activate FX (26). The regions in the light chain implicated in mediating platelet binding are the loop (Gly 4 -Gln 11 ) within the Gla domain (15,20) and the disulfide-constrained loops 1 and 2 within the EGF2 domain (Cys 88 -Cys 109 ) (2)(3)(4). The residues contained within these regions appear to be both necessary and sufficient for mediating the incorporation of the enzyme, FIXa, into the FX-activating complex on the platelet surface (6), because the impaired FXa generation rates resulted exclusively from increases in FIXa K d(app) values rather than K d(app)FVIIIa or K m(app) values.